Known-Plaintext Attacks to Thumbnail-Preservation Encryption Using Pix2pix Generative Adversarial Network
Zhiyang Li, Dong Xie, Sanqiang Liu, Fulong Chen, Peng Hu, Taochun Wang
Abstract
General image encryption schemes transform plaintext images into snowflake-like ciphertext patterns through efficient cryptographic permutation and confusion primitives. However, these encryption schemes cannot achieve the goal of privacy computing that data are available but not visible. Thumbnail-preserving encryption (TPE) not only ensures the privacy of ciphertext images, but also allows legitimate users of cloud servers to search through thumbnails in the ciphertext image space. But as far as we know, there has not been a detailed security analysis of TPE schemes up to now.In this paper, we propose a universal attack framework, called AttackTPE, based on pix2pix generative adversarial network under known-plaintext attack (KPA) model. The experimental results demonstrate that the average structural similarity (SSIM) between decrypted images and plaintext images of different block sizes ranges from 0.7024 to 0.8219, and the average peak signal-to-noise ratio (PSNR) between them ranges from 18.6825 to 24.2575. Additionally, the smaller the block size of ciphertext images, the higher the SSIM and PSNR.
BibTeX
@inproceedings{icassp2025_knownplaintextat,
title = {Known-Plaintext Attacks to Thumbnail-Preservation Encryption Using Pix2pix Generative Adversarial Network},
author = {Zhiyang Li and Dong Xie and Sanqiang Liu and Fulong Chen and Peng Hu and Taochun Wang},
booktitle = {ICASSP 2025},
year = {2025}
}